Siddharth Kotian

Experience

• 

  RUGVED Systems

  Dec 2020 - Jun 2023

  - Developed real-time algorithms for lane and object detection for UGVs. Collaborated with my AI team members to optimize performance and reliability.- Co-created software for dynamic hand gesture recognition, achieving an accuracy greater than 95%. Designed and implemented a novel algorithm using CNN and LSTM for more accurate and efficient recognition.- Co-created an intelligent gun system using raspberry pi and stepper motors, which automatically detects human targets using YOLO and fires. Collaborated with mechanical and electronic subsystems to integrate hardware and software components and optimize performance.- Completed relevant coursework in the fields of robotics, computer vision, and deep learning and participated in multiple college-level and national competitions like MANTHAN, Smart India Hackathon, IIT-Bombay Techfest, KAVACH Hackathon. Show less

  • Senior Member - AI subsystem

    Dec 2020 - Jun 2023

  • AI Taskphase Recruit

    Dec 2020 - Jun 2021
• 

  IEEE Student Branch Manipal

  Dec 2021 - Oct 2022

  Management Committee Member
• 

  Infyuva Tech Solutions Pvt Ltd

  Jul 2022 - Dec 2023

  Computer Vision Engineer

  - As an intern my primary role was to develop the existing research paper models into a working prototype which can be deployed on an embedded system. - Created an iris detection and tracking software using open sourced Computer vision libraries. Thoroughly tested the working of the model on the Rasberry Pi system.- Integrated the iris tracking software with stepper motors via the Rasberry Pi- Finetuned and improved upon the existing segmentation and classification models.- Co-created a cross-platform GUI which demonstrated the working of the software product Show less

• 

  Google Crowdsource

  Jan 2023 - now

  ExploreML with Google Crowdsource Facilitator
• 

  Indian Institute of Science (IISc)

  Jun 2023 - Aug 2023

  Research Intern

  - Developed and tested semantic segmentation models for lesion segmentation in diabetic retinopathy.- Leveraged state-of-the-art segmentation models, including SAM, MedSAM, DeepLab, and U-Net, to achieve high-precision segmentation.- Used transfer learning for multi-class classification to grade diabetic retinopathy severity.- Achieved near state-of-the-art model performances with classification accuracies exceeding 90% and AUC-ROC scores exceeding 0.98.- Improved the effectiveness and reliability of models used in diabetic retinopathy detection and severity grading. Show less

• 

  BlackRock

  Jan 2024 - now

  • Analyst

    Jul 2024 - now

  • Data Engineer Intern

    Jan 2024 - Jun 2024